TheoremBase

Bisection without choice: assuming an open cover of Q with no finite subcover, a recursion on the natural numbers halves every side of the current box in turn, always keeping the first half with no finite subcover; the suprema of the left endpoints give a point of every box, and one member of the cover contains a ball about it that swallows a late box, a contradiction.

Proof

Each result cited is universally quantified over the data in its own statement. No choice principle is used: the boxes below are defined by recursion from maps given by formulas, and every other object is obtained from an existential statement or defined by a formula.

Let nn, dEd_{E}, aa, bb and QQ be as in the statement. By Rn\mathbb{R}^n with the Euclidean Distance Is a Complete Metric Space with Coordinatewise Convergence §metric, dEd_{E} is a metric, so it satisfies Metric Space §symmetric; TdE\mathcal{T}_{d_{E}} is the topology of Metric Spaces: Balls, the Metric Topology, and the Real and Complex Numbers as Metric Spaces §topology, and TQ\mathcal{T}_{Q} is the subspace topology on QQ. By Compact Topological Space and Compact Subset §compact-subset and Compact Topological Space and Compact Subset §compact we must show: for every subset U\mathcal{U} of TQ\mathcal{T}_{Q} with Q⊆⋃UQ\subseteq\bigcup\mathcal{U} there is a finite subset V\mathcal{V} of U\mathcal{U} with Q⊆⋃VQ\subseteq\bigcup\mathcal{V}. Fix such a U\mathcal{U}.

Small sets. Call a subset SS of QQ small if there is a finite subset V\mathcal{V} of U\mathcal{U} with S⊆⋃VS\subseteq\bigcup\mathcal{V}. This is a predicative formula in SS with parameter U\mathcal{U} (it quantifies over sets only: subsets of U\mathcal{U}, natural numbers and maps), so it may be used to form sets by Sets and Maps: Ordinary Notation §set-builder. We must show that QQ is small; suppose, for a contradiction, that QQ is not small.

(F1) If S1,S2⊆QS_{1},S_{2}\subseteq Q are small, so is S1∪S2S_{1}\cup S_{2}: if V1,V2\mathcal{V}_{1},\mathcal{V}_{2} are finite subsets of U\mathcal{U} with Si⊆⋃ViS_{i}\subseteq\bigcup\mathcal{V}_{i}, then V1∪V2\mathcal{V}_{1}\cup\mathcal{V}_{2} is a finite subset of U\mathcal{U} by Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice §union, and S1∪S2⊆⋃(V1∪V2)S_{1}\cup S_{2}\subseteq\bigcup(\mathcal{V}_{1}\cup\mathcal{V}_{2}).

Boxes. Points of Rn\mathbb{R}^{n} are maps [n]→R[n]\to\mathbb{R} with components as in Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations §membership, and two points are equal when their components agree, by Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations §equality. Let

D={(α,β)∈Rn×Rn:ak≤αk≤βk≤bk for every k∈[n]},D=\{(\alpha,\beta)\in\mathbb{R}^{n}\times\mathbb{R}^{n}:a_{k}\le\alpha_{k}\le\beta_{k}\le b_{k}\text{ for every }k\in[n]\},

and for p=(α,β)∈Dp=(\alpha,\beta)\in D let Q(p)={x∈Rn:αk≤xk≤βk for every k∈[n]}Q(p)=\{x\in\mathbb{R}^{n}:\alpha_{k}\le x_{k}\le\beta_{k}\text{ for every }k\in[n]\}, a subset of QQ (the conditions defining DD and Q(p)Q(p) being predicative by Predicative Substitution of Values of Functions and Binary Operations, and Predicative Formulas for Negatives, Differences, Absolute Values, Distances, Reciprocals and Quotients in an Ordered Field §value-substitution); note (a,b)∈D(a,b)\in D and Q((a,b))=QQ((a,b))=Q. For p=(α,β)p=(\alpha,\beta) and p′=(α′,β′)p'=(\alpha',\beta') in DD write p′⊑pp'\sqsubseteq p if αk≤αk′\alpha_{k}\le\alpha'_{k} and βk′≤βk\beta'_{k}\le\beta_{k} for every k∈[n]k\in[n]; then Q(p′)⊆Q(p)Q(p')\subseteq Q(p), and ⊑\sqsubseteq is transitive. For p=(α,β)∈Dp=(\alpha,\beta)\in D and k∈[n]k\in[n] write wk(p)=βk−αk≥0w_{k}(p)=\beta_{k}-\alpha_{k}\ge0 for the kk-th side length, a real number by Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §negative.

Halving one side. Let j∈[n]j\in[n] and p=(α,β)∈Dp=(\alpha,\beta)\in D, and let c=(αj+βj)/2c=(\alpha_{j}+\beta_{j})/2, a real number by Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §halves, so αj≤c≤βj\alpha_{j}\le c\le\beta_{j}. Let β′∈Rn\beta'\in\mathbb{R}^{n} have components βj′=c\beta'_{j}=c and βk′=βk\beta'_{k}=\beta_{k} for k≠jk\neq j, and α′∈Rn\alpha'\in\mathbb{R}^{n} components αj′=c\alpha'_{j}=c and αk′=αk\alpha'_{k}=\alpha_{k} for k≠jk\neq j (maps on [n][n] given by Sets and Maps: Ordinary Notation §maps, their conditions being predicative by Predicative Substitution of Values of Functions and Binary Operations, and Predicative Formulas for Negatives, Differences, Absolute Values, Distances, Reciprocals and Quotients in an Ordered Field §value-substitution). Put λj(p)=(α,β′)\lambda_{j}(p)=(\alpha,\beta') and ρj(p)=(α′,β)\rho_{j}(p)=(\alpha',\beta). Both lie in DD, both satisfy q⊑pq\sqsubseteq p, and for q∈{λj(p),ρj(p)}q\in\{\lambda_{j}(p),\rho_{j}(p)\},

wj(q)=12wj(p),wk(q)=wk(p) for k∈[n], k≠j,w_{j}(q)=\tfrac{1}{2}w_{j}(p),\qquad w_{k}(q)=w_{k}(p)\text{ for }k\in[n],\ k\neq j,

since c−αj=βj−c=(βj−αj)/2c-\alpha_{j}=\beta_{j}-c=(\beta_{j}-\alpha_{j})/2. Moreover Q(p)=Q(λj(p))∪Q(ρj(p))Q(p)=Q(\lambda_{j}(p))\cup Q(\rho_{j}(p)): the inclusion ⊇\supseteq holds because both are ⊑p\sqsubseteq p, and for x∈Q(p)x\in Q(p) either xj≤cx_{j}\le c, and then x∈Q(λj(p))x\in Q(\lambda_{j}(p)), or c<xjc<x_{j}, and then x∈Q(ρj(p))x\in Q(\rho_{j}(p)).

Let φ:[n]×D→D\varphi:[n]\times D\to D be the set

φ={((j,p),q)∈([n]×D)×D: (Q(λj(p)) is not small and q=λj(p)) or (Q(λj(p)) is small and q=ρj(p))},\varphi=\{((j,p),q)\in([n]\times D)\times D:\ (Q(\lambda_{j}(p))\text{ is not small and }q=\lambda_{j}(p))\text{ or }(Q(\lambda_{j}(p))\text{ is small and }q=\rho_{j}(p))\},

formed by Sets and Maps: Ordinary Notation §set-builder; exactly one of the two cases holds for each (j,p)(j,p), so φ\varphi is a map [n]×D→D[n]\times D\to D by Functions, Values of a Function, and Functions from One Class to Another §map. Write φj(p)=φ(j,p)\varphi_{j}(p)=\varphi(j,p); thus φj(p)\varphi_{j}(p) is the left half λj(p)\lambda_{j}(p) if that is not small, and the right half ρj(p)\rho_{j}(p) otherwise.

(F2) If Q(p)Q(p) is not small, then Q(φj(p))Q(\varphi_{j}(p)) is not small. If Q(λj(p))Q(\lambda_{j}(p)) is not small, φj(p)=λj(p)\varphi_{j}(p)=\lambda_{j}(p). Otherwise Q(λj(p))Q(\lambda_{j}(p)) is small and φj(p)=ρj(p)\varphi_{j}(p)=\rho_{j}(p); were Q(ρj(p))Q(\rho_{j}(p)) small too, Q(p)=Q(λj(p))∪Q(ρj(p))Q(p)=Q(\lambda_{j}(p))\cup Q(\rho_{j}(p)) would be small by (F1).

One round: halving every side. Let g:N×D→Dg:\mathbb{N}\times D\to D be the map with g(k,u)=φk(u)g(k,u)=\varphi_{k}(u) for k∈[n]k\in[n] and g(k,u)=ug(k,u)=u for k∈Nk\in\mathbb{N} with n<kn<k, a set by Sets and Maps: Ordinary Notation §set-builder and a map since exactly one case applies to each (k,u)(k,u). For p∈Dp\in D, by Recursion on the Natural Numbers Starting at One §recursion (with the set DD, the element pp and the map gg) there is exactly one map hp:N→Dh_{p}:\mathbb{N}\to D with hp(1)=ph_{p}(1)=p and hp(k+1)=g(k,hp(k))h_{p}(k+1)=g(k,h_{p}(k)) for every k∈Nk\in\mathbb{N}. Let Φ:D→D\Phi:D\to D be the set of pairs (p,q)∈D×D(p,q)\in D\times D such that q=h(n+1)q=h(n+1) for some map h:N→Dh:\mathbb{N}\to D with h(1)=ph(1)=p and h(k+1)=g(k,h(k))h(k+1)=g(k,h(k)) for every k∈Nk\in\mathbb{N}; it is formed by Sets and Maps: Ordinary Notation §set-builder (the formula quantifies over the set DND^{\mathbb{N}} only), and it is a map with Φ(p)=hp(n+1)\Phi(p)=h_{p}(n+1) by the existence and uniqueness of hph_{p}.

(F3) Let p=(α,β)∈Dp=(\alpha,\beta)\in D. For every k∈[n+1]k\in[n+1]: hp(k)⊑ph_{p}(k)\sqsubseteq p; Q(hp(k))Q(h_{p}(k)) is not small if Q(p)Q(p) is not small; and for i∈[n]i\in[n], wi(hp(k))=12wi(p)w_{i}(h_{p}(k))=\frac{1}{2}w_{i}(p) if i<ki<k and wi(hp(k))=wi(p)w_{i}(h_{p}(k))=w_{i}(p) if k≤ik\le i. By induction on kk, The Natural Numbers and the Natural Numbers with Zero: Arithmetic, Order, Induction and Recursion §induction, applied to the class {k∈N:k≤n+1⇒P(k)}\{k\in\mathbb{N}:k\le n+1\Rightarrow P(k)\}, where P(k)P(k) is the claim just stated for kk; this class is formed by a predicative formula with set parameters only (among them pp, hph_{p}, nn and U\mathcal{U}). For k=1k=1, hp(1)=ph_{p}(1)=p and there is no i∈[n]i\in[n] with i<1i<1. Let k∈[n]k\in[n] and suppose the claim for kk. Then hp(k+1)=g(k,hp(k))=φk(hp(k))h_{p}(k+1)=g(k,h_{p}(k))=\varphi_{k}(h_{p}(k)), which is λk(hp(k))\lambda_{k}(h_{p}(k)) or ρk(hp(k))\rho_{k}(h_{p}(k)). Hence hp(k+1)⊑hp(k)⊑ph_{p}(k+1)\sqsubseteq h_{p}(k)\sqsubseteq p; non-smallness passes on by (F2); and for i∈[n]i\in[n]: if i<ki<k, then i≠ki\neq k and wi(hp(k+1))=wi(hp(k))=12wi(p)w_{i}(h_{p}(k+1))=w_{i}(h_{p}(k))=\frac{1}{2}w_{i}(p); if i=ki=k, then wk(hp(k+1))=12wk(hp(k))=12wk(p)w_{k}(h_{p}(k+1))=\frac{1}{2}w_{k}(h_{p}(k))=\frac{1}{2}w_{k}(p); if k+1≤ik+1\le i, then i≠ki\neq k and wi(hp(k+1))=wi(hp(k))=wi(p)w_{i}(h_{p}(k+1))=w_{i}(h_{p}(k))=w_{i}(p). Since i<k+1i<k+1 exactly when i<ki<k or i=ki=k, this is the claim for k+1k+1.

Taking k=n+1k=n+1, every i∈[n]i\in[n] satisfies i<n+1i<n+1, so for every p∈Dp\in D: Φ(p)⊑p\Phi(p)\sqsubseteq p, wi(Φ(p))=12wi(p)w_{i}(\Phi(p))=\frac{1}{2}w_{i}(p) for every i∈[n]i\in[n], and Q(Φ(p))Q(\Phi(p)) is not small if Q(p)Q(p) is not small.

The nested boxes. By Recursion on the Natural Numbers Starting at One §recursion (with the set DD, the element (a,b)(a,b) and the map (m,u)↦Φ(u)(m,u)\mapsto\Phi(u) on N×D\mathbb{N}\times D, given by Sets and Maps: Ordinary Notation §maps and Predicative Substitution of Values of Functions and Binary Operations, and Predicative Formulas for Negatives, Differences, Absolute Values, Distances, Reciprocals and Quotients in an Ordered Field §value-substitution) there is a map F:N→DF:\mathbb{N}\to D with F(1)=(a,b)F(1)=(a,b) and F(m+1)=Φ(F(m))F(m+1)=\Phi(F(m)) for every m∈Nm\in\mathbb{N}. Write F(m)=(α(m),β(m))F(m)=(\alpha^{(m)},\beta^{(m)}). By induction on mm, The Natural Numbers and the Natural Numbers with Zero: Arithmetic, Order, Induction and Recursion §induction, applied to the class of those m∈Nm\in\mathbb{N} for which (N1)--(N3) below hold for every k∈[n]k\in[n], formed by a predicative formula with set parameters only (among them FF, aa, bb, nn and U\mathcal{U}), and using the last paragraph, for every m∈Nm\in\mathbb{N} and k∈[n]k\in[n]: (N1) Q(F(m))Q(F(m)) is not small (for m=1m=1 this is the assumption on QQ); (N2) βk(m)−αk(m)=(bk−ak)(1/2)m−1\beta^{(m)}_{k}-\alpha^{(m)}_{k}=(b_{k}-a_{k})(1/2)^{m-1} (exponents lie in N0\mathbb{N}_{0}, m−1m-1 being the difference in N0\mathbb{N}_{0} by Commutative Rings, Fields and Ordered Fields: Standard Notation §numerals, and (1/2)0=1(1/2)^{0}=1 by Commutative Rings, Fields and Ordered Fields: Standard Notation §rings and Powers with Exponents in the Natural Numbers with Zero §zero; 1/21/2 and its powers lie in R\mathbb{R} by Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §two, Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §quotient and Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §powers); (N3) F(m+1)⊑F(m)F(m+1)\sqsubseteq F(m), that is αk(m)≤αk(m+1)\alpha^{(m)}_{k}\le\alpha^{(m+1)}_{k} and βk(m+1)≤βk(m)\beta^{(m+1)}_{k}\le\beta^{(m)}_{k}. By (N3), for each k∈[n]k\in[n] the sequence m↦αk(m)m\mapsto\alpha^{(m)}_{k} is nondecreasing and m↦βk(m)m\mapsto\beta^{(m)}_{k} is nonincreasing, so by Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §monotone, αk(m)≤αk(m′)\alpha^{(m)}_{k}\le\alpha^{(m')}_{k} and βk(m′)≤βk(m)\beta^{(m')}_{k}\le\beta^{(m)}_{k} whenever m≤m′m\le m'.

The limit point. Let k∈[n]k\in[n]. For all m,m′∈Nm,m'\in\mathbb{N}, αk(m)≤βk(m′)\alpha^{(m)}_{k}\le\beta^{(m')}_{k}: if m≤m′m\le m', then αk(m)≤αk(m′)≤βk(m′)\alpha^{(m)}_{k}\le\alpha^{(m')}_{k}\le\beta^{(m')}_{k}; if m′<mm'<m, then αk(m)≤βk(m)≤βk(m′)\alpha^{(m)}_{k}\le\beta^{(m)}_{k}\le\beta^{(m')}_{k}. So the nonempty set Ak={αk(m):m∈N}⊆RA_{k}=\{\alpha^{(m)}_{k}:m\in\mathbb{N}\}\subseteq\mathbb{R} is bounded above, and it has a supremum sk=sup⁡Aks_{k}=\sup A_{k} by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §supremum, unique as the least upper bound. Being an upper bound of AkA_{k}, sk≥αk(m)s_{k}\ge\alpha^{(m)}_{k}; being the least one, sk≤βk(m)s_{k}\le\beta^{(m)}_{k}, for every m∈Nm\in\mathbb{N}. Let s∈Rns\in\mathbb{R}^{n} be the map k↦skk\mapsto s_{k} on [n][n], given by Sets and Maps: Ordinary Notation §maps and Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations §membership. Then s∈Q(F(m))s\in Q(F(m)) for every m∈Nm\in\mathbb{N}; in particular s∈Q(F(1))=Qs\in Q(F(1))=Q.

The contradiction. Since s∈Q⊆⋃Us\in Q\subseteq\bigcup\mathcal{U}, there is W∈UW\in\mathcal{U} with s∈Ws\in W. As W∈TQW\in\mathcal{T}_{Q}, W=Q∩UW=Q\cap U for some U∈TdEU\in\mathcal{T}_{d_{E}} by Subspace Topology §subspace-topology, and UU is open in (Rn,dE)(\mathbb{R}^{n},d_{E}) by Metric Spaces: Balls, the Metric Topology, and the Real and Complex Numbers as Metric Spaces §topology. As s∈Us\in U, there is a real r>0r>0 with BdE(s,r)⊆UB_{d_{E}}(s,r)\subseteq U, by Open Subset of a Metric Space §open.

Let w=∥b−a∥≥0w=\lVert b-a\rVert\ge0, the Euclidean norm of b−ab-a. For k∈[n]k\in[n], (b−a)k=bk−ak≥0(b-a)_{k}=b_{k}-a_{k}\ge0 by Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations §operations, so bk−ak=∣(b−a)k∣≤wb_{k}-a_{k}=|(b-a)_{k}|\le w by Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square §coordinate. Now, in this order: rr is fixed above; put ε=r/(2(nw+1))>0\varepsilon=r/(2(nw+1))>0, a real number by Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §naturals, Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §two and Closure Facts in an Ordered Field: Negatives, Differences, Quotients, Absolute Values, Powers, Natural Numbers, Halves and Minima §quotient; by Completeness of the Real Numbers for Sequences: Monotone Convergence, the Bolzano-Weierstrass Theorem and Cauchy Sequences §geometric with ratio 1/21/2, (1/2)m→0(1/2)^{m}\to0, so by Convergent Sequences of Real Numbers §converges there is N∈NN\in\mathbb{N} with (1/2)N=∣(1/2)N−0∣<ε(1/2)^{N}=|(1/2)^{N}-0|<\varepsilon. Then (1/2)N−1=2(1/2)N<r/(nw+1)(1/2)^{N-1}=2(1/2)^{N}<r/(nw+1).

Let x∈Q(F(N))x\in Q(F(N)) and y=x−sy=x-s, so yk=xk−sky_{k}=x_{k}-s_{k} for k∈[n]k\in[n] by Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations §operations. Both xkx_{k} and sks_{k} lie between αk(N)\alpha^{(N)}_{k} and βk(N)\beta^{(N)}_{k}, so by (N2)

∣yk∣≤βk(N)−αk(N)=(bk−ak)(1/2)N−1≤w(1/2)N−1for every k∈[n].|y_{k}|\le\beta^{(N)}_{k}-\alpha^{(N)}_{k}=(b_{k}-a_{k})(1/2)^{N-1}\le w(1/2)^{N-1}\quad\text{for every }k\in[n].

By Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square §coordinate, ∥y∥≤nw(1/2)N−1≤(nw+1)(1/2)N−1<r\lVert y\rVert\le nw(1/2)^{N-1}\le(nw+1)(1/2)^{N-1}<r. By Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square §distance and symmetry of dEd_{E}, dE(s,x)=dE(x,s)=∥x−s∥<rd_{E}(s,x)=d_{E}(x,s)=\lVert x-s\rVert<r, so x∈BdE(s,r)x\in B_{d_{E}}(s,r) by Open Ball in a Metric Space §membership, hence x∈Ux\in U. As also x∈Qx\in Q, x∈Q∩U=Wx\in Q\cap U=W. Thus Q(F(N))⊆W=⋃{W}Q(F(N))\subseteq W=\bigcup\{W\}, and {W}\{W\} is a finite subset of U\mathcal{U} by Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice §small, so Q(F(N))Q(F(N)) is small. This contradicts (N1).

Hence QQ is small: there is a finite subset V\mathcal{V} of U\mathcal{U} with Q⊆⋃VQ\subseteq\bigcup\mathcal{V}. As U\mathcal{U} was arbitrary, (Q,TQ)(Q,\mathcal{T}_{Q}) is compact by Compact Topological Space and Compact Subset §compact, and QQ is compact in (Rn,TdE)(\mathbb{R}^{n},\mathcal{T}_{d_{E}}).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…